Z. Chen, B. Abidi, D. Page, and M.
Abidi, "Gray Level Grouping (GLG): An Automatic Method for Optimized Image
230 Contrast Enhancement - Part I: The Basic Method," IEEE Trans. on Image Processing, Vol. 15, No. 8, pp. 2290 -
2302, August 2006.
2290
Gray-Level Grouping (GLG): An Automatic Method
for Optimized Image Contrast Enhancement—Part I:
The Basic Method
ZhiYu Chen, Senior Member, IEEE, Besma R. Abidi, Senior Member, IEEE, David L. Page, Member, IEEE, and
Mongi A. Abidi, Member, IEEE
Abstract—Contrast enhancement has an important role in
image processing applications. Conventional contrast enhance-
ment techniques either often fail to produce satisfactory results for
a broad variety of low-contrast images, or cannot be automatically
applied to different images, because their parameters must be
specified manually to produce a satisfactory result for a given
image. This paper describes a new automatic method for contrast
enhancement. The basic procedure is to first group the histogram
components of a low-contrast image into a proper number of
bins according to a selected criterion, then redistribute these bins
uniformly over the grayscale, and finally ungroup the previously
grouped gray-levels. Accordingly, this new technique is named
gray-level grouping (GLG). GLG not only produces results supe-
rior to conventional contrast enhancement techniques, but is also
fully automatic in most circumstances, and is applicable to a broad
variety of images. An extension of GLG, selective GLG (SGLG),
and its variations will be discussed in Part II of this paper. SGLG
selectively groups and ungroups histogram components to achieve
specific application purposes, such as eliminating background
noise, enhancing a specific segment of the histogram, and so on.
The extension of GLG to color images will also be discussed in
Part II.
Index Terms—Contrast enhancement, gray-level grouping, his-
togram, quality measure.
I. INTRODUCTION AND RELATED WORK
ONTRAST enhancement has an important role in image
C processing applications. Numerous contrast enhancement
techniques exist in literature, such as gray-level transformation
based techniques (e.g., logarithm transformation, power-law
transformation, piecewise-linear transformation, etc.) and
histogram processing techniques (e.g., histogram equalization
(HE), histogram specification, etc.) [1]. Conventional contrast
enhancement techniques generally yield satisfactory results if
the proper technique is selected for a given application along
with the proper processing parameters. However, conventional
Manuscript received February 1, 2005; revised August 26, 2005. This
work was supported in part by the DOE University Research Program
in Robotics under Grant DOE-DE-FG02-86NE37968, in part by the
DOD/TACOM/NAC/ARC Program R01-1344-18, in part by the FAA/NSSA
Program R01-1344-48/49, and in part by the Office of Naval Research under
Grant N000143010022. The associate editor coordinating the review of this
manuscript and approving it for publication was Dr. Hassan Foroosh.
The authors are with the Electrical and Computer Engineering Depart-
ment, University of Tennessee, Knoxville, TN 37996-2100 USA (e-mail:
zychen@utk.edu; besma@utk.edu; dpage@utk.edu; abidi@utk.edu).
Digital Object Identifier 10.1109/TIP.2006.875204
1057-7149/$20.00 © 2006 IEEE
IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 8, AUGUST 2006
contrast enhancement techniques often fail in producing satis-
factory results for a broad range of low-contrast images, such
as images characterized by the fact that the amplitudes of their
histogram components are very high at one or several locations
on the grayscale, while they are very small, but not zero, in
the rest of the grayscale. This makes it difficult to increase the
image contrast by simply stretching its histogram or by using
simple gray-level transformations. The high amplitude of the
histogram components corresponding to the image background
also often prevents the use of the HE techniques, which could
cause a washed-out effect on the appearance of the output
image and/or amplify the background noise. Figs. 1 and 4 show
examples of low-contrast images and the results of treating
them with conventional contrast enhancement techniques.
Fig. 1(a) shows an original low-contrast image of the Mars
moon, Phobos, and Fig. 2(a) its histogram. Fig. 1(b) is the re-
sult of HE, exhibiting a washed-out appearance which is not
acceptable for many applications. The cause for the washed-out
appearance is that the left half of the grayscale on the histogram
of the equalized image is simply empty, as shown in Fig. 2(b).
Fig. 1(c) is the resulting image of histogram specification, and
Fig. 2(c) its histogram, which is better than the HE result, but
still has an unsatisfactory appearance. More importantly, one
major disadvantage of the histogram specification technique is
that the desired histogram of the resulting image has to be spec-
ified manually, and this precludes the technique from being ap-
plied automatically. The manually specified desired histogram
used in the treatment is depicted in Fig. 3 [1]. Fig. 4(a) shows a
low-contrast X-ray image of luggage. Its HE result in Fig. 4(b)
also has a washed-out look.
Numerous advanced histogram-based contrast enhancement
techniques have been developed, but most of them are deriva-
tives of conventional techniques (e.g., HE, etc.) [2]–[25], such as
bi-histogram equalization (BHE), block-overlapped histogram
equalization, multi-scale adaptive histogram equalization, shape
preserving local histogram modification, and so on. The mean
brightness of histogram-equalized image is always the middle
gray-level regardless of the input mean, and this is undesirable
in certain applications where brightness preservation is neces-
sary. This characteristic of HE may also lead to a washed-out
appearance, amplified noise or other annoying artifacts in the
resulting image. BHE was proposed to preserve the brightness
by separating the input image’s histogram into two parts based
on its mean—one ranges from the minimum gray level to the
mean gray level, the other from the mean to the maximum. The
CHEN et al.: GLG: AN AUTOMATIC METHOD FOR OPTIMIZED IMAGE CONTRAST ENHANCEMENT—PART I
Fig. 1. Mars moon—Phobos. (a) Low-contrast original image. (b) Result of
HE, which has a washed-out appearance. (c) Result of histogram specification,
though better than the HE result, still has an unsatisfactory appearance. More-
over, this technique is not automatic; the desired histogram profile must be man-
ually specified. (d) Result of gray-level grouping, has a crisper look. The result
is produced fully automatically (P ixDist and T EN are quality measures that
will be discussed in Section III). (Original image courtesy of Dr. R. C. Gonzalez
[1]).
two histograms are then equalized independently [2]. Equal area
dualistic subimage histogram equalization (DSIHE) is similar to
BHE except that DSIHE separates the histogram at the median
gray level—the gray level with cumulative probability equal to
0.5 instead of the mean [3]. These two techniques usually out-
perform the basic HE technique. However, they have the same
limitations of HE and cannot enhance some images well, as they
still perform the HE operation in each grayscale segment, just
limiting the drawbacks of HE within each grayscale segment.
The global HE method cannot adapt to local brightness fea-
tures of the input image because it uses histogram information
over the whole image. This fact limits the contrast-stretching
ratio in some parts of the image, and causes significant contrast
losses in the background and other small regions. To overcome
this limitation, some local histogram-equalization methods
2291
Fig. 2. Histograms of the images of Phobos in Fig. 1. (a) Histogram of the
low-contrast original image. (b) Result of HE. Nearly half of the grayscale is
wasted. (c) Result of histogram specification. The desired histogram profile must
be manually specified. (d) Result of gray-level grouping. The grayscale has been
utilized fully and efficiently (the leftmost component in the histograms is the
2
largest peak whose actual amplitude is 3:67 10 . It is truncated so that the
rest of the histograms can be displayed on a proper scale).
Fig. 3. Manually specified desired histogram profile used to produce the his-
togram specification result in Fig. 1(c).
have been developed. A natural extension of global histogram
equalization is termed adaptive histogram equalization (AHE),
which divides the input image into an array of subimages, each
subimage is histogram-equalized independently, and then the
processed subimages are fused together with bilinear interpo-
lation [4].
Another local method is called block-overlapped histogram
equalization [5], in which a rectangular subimage of the input
image is first defined, a histogram of that block is obtained,
and then its histogram-equalization function is determined.
Thereafter, the center pixel of the block is histogram equalized
using this function. The center of the rectangular block is then
moved to the adjacent pixel and the histogram equalization
is repeated. This procedure is repeated pixel by pixel for all
input pixels. Since local histogram equalization must be per-
formed for all pixels in the entire image, the computational
complexity of this method is very high. Instead of using rect-
angular blocks, shape preserving local histogram modification
employs connected components and level-sets for contrast
2292
Fig. 4. X-ray image of luggage. (a) Low-contrast original image. (b) Result of
HE, which has an unsatisfactory appearance. (c) Result of gray-level grouping,
has a sharper look. The result is produced fully automatically.
enhancement [6]. Multiscale adaptive histogram equalization
[11] and other multiscale contrast enhancement techniques
[7], [22] use multiscale analysis to decompose the image into
subbands, and apply corresponding enhancement techniques to
the high-frequency subband, and then combine the enhanced
high-frequency subband with the low-frequency subband to
reconstruct the output image.
The above mentioned advanced contrast enhancement tech-
niques usually outperform conventional techniques. However,
they still have limitations and cannot handle certain classes of
images well and/or are not fully automatic methods.
Our motivation is to develop a new contrast enhancement
technique which not only produces better results, but is also
general and can be automatically applied to a broad variety
of images. This paper introduces a new histogram-based
optimized contrast enhancement technique called gray-level
grouping (GLG). The basic procedure of this technique is to
first group the histogram components of a low-contrast image
IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 8, AUGUST 2006
Fig. 5. Histograms of a virtual low-contrast image treated by different enhance-
ment methods. (a) Original. (b) Result of histogram equalization. Half of the
grayscale is wasted. (c) Result of linear contrast stretching. Contrast enhance-
ment is not strong for histogram components which are originally very close to
each other. (d) Optimal histogram of the enhanced image.
into a proper number of groups according to a certain crite-
rion, then redistribute these groups of histogram components
uniformly over the grayscale so that each group occupies a
grayscale segment of the same size as the other groups, and
finally ungroup the previously grouped gray-levels.
In the next section, the principle and algorithm of the gray-
level grouping technique will be described. The computational
complexity of this method and comparison with a few of con-
ventional methods using a benchmark quality measure will be
discussed in Section III. An extension of the basic GLG to local
contrast enhancement approaches, adaptive gray-level grouping
(A-GLG), will be presented in Section IV.
II. BASIC GRAY-LEVEL GROUPING (GLG)
Although piecewise-based contrast stretching or histogram
specification might be able to yield satisfactory results if the
proper processing parameters are selected for the image to be
enhanced, they are not general techniques and cannot be auto-
matically applied to other images, since the histogram profile
varies from image to image.
Before introducing the new technique of gray-level grouping,
we first revisit several conventional contrast enhancement tech-
niques, analyze their shortcomings, and try to overcome them
in developing the new method.
Fig. 5(a) illustrates the histogram of a virtual low-contrast
image. The histogram contains four components, which are
clustered in the central part of the grayscale. The amplitude
CHEN et al.: GLG: AN AUTOMATIC METHOD FOR OPTIMIZED IMAGE CONTRAST ENHANCEMENT—PART I
of the second component is half that of the leftmost compo-
nent, and the right two components are half the second one.
Fig. 5(b) shows the result of performing histogram equalization
on Fig. 5(a). Due to the highest amplitude of the leftmost
component, the resulting histogram is shifted toward the right
side of the grayscale. The left half of the grayscale is empty and
this produces a washed-out appearance in the output image.
The objective of histogram equalization is to achieve a uniform
histogram. However, this can be achieved only on continuous
histograms. For digital images and their discrete histograms,
histogram equalization simply cannot redistribute the histogram
components uniformly in most cases, because of the continuous
nature of the technique.
Fig. 5(c) shows the result of performing linear contrast stretch
on Fig. 5(a). The resulting histogram spreads over the full range
of grayscale, so there is no washed-out appearance in the output
image. However, it can be noted that the right two histogram
components are still quite close to each other, so are the left
two. As a result, the contrast enhancement in some regions of
the output image is not the strongest. Since the left two his-
togram components are far away from the right two, they might
be over-contrasted in the output image. Therefore, the contrast
enhancement in the resulting image might also be unbalanced.
Furthermore, linear contrast stretching is generally not an auto-
matic method, since a piecewise-linear transformation function
need to be manually specified to achieve satisfactory results.
In order to overcome the above shortcomings, the com-
ponents of the desired histogram of the optimal contrast
enhancement result should spread over the full range of the
grayscale, with the bins being away from each other as far as
possible. Fig. 5(d) shows the desired histogram which produces
the strongest contrast enhancement. It can be noted that the four
histogram components are uniformly spread over the entire
grayscale, and are evenly spaced from each other.
The objectives of developing this new technique are as fol-
lows.
• Like histogram equalization, the basic objective of the new
technique is still to achieve a uniform histogram, but for
discrete histograms, in the sense that the histogram com-
ponents are redistributed uniformly over the grayscale.
• Utilize the grayscale more efficiently; conventional con-
trast enhancement techniques such as histogram equal-
ization are likely to leave too much empty space on the
grayscale and cause under or over-contrast.
• Spread the components of histogram over the grayscale in
a controllable and/or efficient way.
• Treat the histogram components in different parts of the
grayscale differently if necessary, in order to satisfy specific
contrast enhancement purposes. This objective will lead to
an extension of the basic GLG technique—selective gray-
level grouping (SGLG), which will be introduced in Part II
of this paper.
• The new technique should be general, and be able to treat
various kinds of images automatically.
The basic principle and procedure of this new technique are
explained below.
• Group the histogram components into a proper number of
gray-level bins according to their amplitudes, in order to
2293
initially reduce the number of gray bins. Therefore, empty
gray levels can be created on the grayscale, allowing
the redistribution of the histogram components in the
next step. Furthermore, this grouping operation results
in a set of gray-level bins whose amplitudes are close to
each other, allowing a quasi-uniform distribution of the
histogram components in the next step.
• Redistribute these groups of histogram components uni-
formly over the grayscale, so that each group occupies a
grayscale segment of the same size as the other groups,
and the concentrated histogram components spread out and
image contrast is increased. The size of the grayscale seg-
ment occupied by each gray-level bin is determined from
the total number of bins. At the same time, the grayscale
is utilized efficiently and the over-contrast problem is also
avoided.
• The histogram components in different segments of the
grayscale can be grouped using different criteria, so they
can be redistributed differently over the grayscale to meet
specific processing purposes, e.g., certain applications may
require different parts of the histogram to be enhanced to
different extents. This step is needed only in SGLG, which
will be discussed in Part II of the paper.
The algorithm of the basic GLG technique is described as
follows, along with a simple example for illustration.
• Let denote the histogram of the original image, with
representing the gray levels on the grayscale
. To perform gray-level grouping, first the nonzero
histogram components are assigned to gray-level bins, or
gray-level groups,
for
(1)
Fig. 6(a) illustrates the histogram of a virtual low-contrast
image, whose gray levels are in the interval [0, 8]. It has
nonzero components and its histogram components
are
and for
After the nonzero histogram components are assigned to
gray-level bins, we have
and
• The left and right limits, and , of the gray-level
interval represented by also need to be recorded. In
this first step, the intervals consist of single values, which
are the gray-level values, , of the original histogram com-
ponents,
for
(2)
2294 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 8, AUGUST 2006

Fig. 6. Illustration of gray-level grouping. (a) Original histogram of a virtual low-contrast image. The bracket indicates the gray levels to be grouped. (b) Histogram
after the first gray-level grouping and ungrouping. (c) Histogram after the second gray-level grouping and ungrouping.

In our example, these parameters are In our example, and . The

bracket in Fig. 6(a) indicates the two histogram compo-
nents to be grouped. The new gray-level bins are

and
• The first occurring smallest is found. The new gray-level intervals represented by the new groups
are

(3)

and is the group index corresponding to the smallest

, i.e., . In our example, and . • Mapping and ungrouping are performed in this step. Now
• Grouping is performed in this step. Group is the total number of gray-level bins has been reduced by
merged with the smaller of its two adjacent neighbors, and one. We can start to construct the transformation function
the gray-level bins are adjusted to create a new set , which maps the gray-level values of pixels in the
of bins , as follows: input image to the desired values in the output image. In
our method, all gray-level bins are redistributed uniformly
for over the entire grayscale, the gray levels are mapped to new
for values, and the combined histogram components are fully
for or partially uncombined. We first calculate the number of
(4) gray levels, , that each gray-level bin will occupy in
where the resulting image. With a total number of bins equal to
, we have

(5)
(9)
and
However, if , this indicates that the
leftmost gray-level bin contains only one gray
for level or one histogram component, which usually corre-
(6)
otherwise. sponds to the background, and it will be matched to gray
level 0 in the resulting image. To prevent this one histogram
The left and right limits of the gray-level intervals repre- component from occupying too many gray levels, we let
sented by also need to be adjusted accordingly

(10)
for
(7)
for where is a constant between 0 and 1, and usually assumes
for a value of 0.8 in our treatments, found through multiple
(8)
for trials to work well with a variety of images.
CHEN et al.: GLG: AN AUTOMATIC METHOD FOR OPTIMIZED IMAGE CONTRAST ENHANCEMENT—PART I 2295

There are four cases to be considered when constructing • By applying the constructed transformation function
. For , we have the fol- to the histogram of the original image, we
lowing. obtain the histogram of the processed image . The
1) If gray-level falls inside gray-level bin , and average distance between pixels on the grayscale,
, this gray level is first mapped is used as a criterion to measure the quality of contrast
onto the right boundary of the gray-level interval as- enhancement. This distance is given by the expression
signed to bin , i.e., , then
it is separated from the group by linear rescaling within
the assigned gray-level interval. Therefore, its trans-
formation function is as shown in (11), at the
bottom of the page. If , constant
for (16)
prevents the background histogram from occupying
too many gray levels. where is the gray-level range of the grayscale,
If , i.e., the bin contains and is the total number of pixels in the image. This
only one gray level, then the transformation function is criterion generally applies only to the gray-level grouping
technique or similar histogram-based techniques, and
for may not be used to judge the quality of images treated
for by other enhancement techniques. A counter example is
(12) given here—if we set the mean gray level of a low-contrast
2) If gray-level falls between gray-level bin image as the threshold, and threshold this image into
and , then its transformation function is a black-and-white image, the average distance between
pixels on the grayscale of the resulting image will be
for the maximum that could be achieved theoretically, how-
for ever, the resulting black-and-white image is obviously
(13) unacceptable for purposes of enhancement. However, the
This ensures that is monotonically increasing GLG process tends to spread the histogram components
along the grayscale, and the gray-level reversal uniformly over the grayscale, preventing the histogram
problem will be avoided in the adaptive approach of components from concentrating in particular locations
the GLG method. on the grayscale. At the same time, a larger will keep
3) the histogram components further away from each other
for better enhancement. Therefore, we consider the av-
erage distance between pixels on the grayscale as a
then (14) sound measure of the quality of images enhanced by GLG
technique, and consider that the maximal corresponds
4)
to the optimal contrast enhancement. Visual evaluations
of multiple images during our testing also confirmed the
then (15) validity of this measure. This quality measure is essen-
tial in the GLG process to achieve the optimal result.
The constructed gray-level transformation function It is worth noting that this image contrast criterion, the
for is stored in computer average distance between pixels on the grayscale, is not
memory. In our example, we let for simplicity and inherent to the GLG algorithm, but could be used in
have . The transformed gray other histogram-based algorithms (especially histogram
levels are equalization) as well. However, we suggest that this cri-
terion be used with caution. The use of this criterion and
a well-known criterion on images in this paper will be
discussed in Section III.
In some cases (e.g., the background occupies a large
All resulting gray levels are rounded to the closest integer, percentage area in the image), in order to achieve the
and the histogram of the resulting image is shown in Fig. optimal result, the gray levels corresponding to the image
6(b). background may be excluded when calculating . For

for
(11)
for
2296 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 8, AUGUST 2006

Fig. 7. Flow chart of the optimized gray-level grouping algorithm.

many images, the histogram components corresponding
to the background are the highest and distinct in the
histogram profile. Therefore, the approximate area of the
background can be calculated automatically by summing
the amplitudes of the histogram components of the back-
ground, given that the algorithm is notified by the user
that the input image has a large-area background. If the
background occupies a percentage area in the image larger
than a user specified threshold (e.g., 40%), the background
gray levels are then excluded when calculating .
In our example, of Fig. 6(a) is 1.36, and of Fig. 6(b)
is 1.87.
To determine the optimal number of gray-level bins that
will lead to the optimal contrast enhancement, we need to
CHEN et al.: GLG: AN AUTOMATIC METHOD FOR OPTIMIZED IMAGE CONTRAST ENHANCEMENT—PART I 2297

Fig. 8. Flow chart of the FGLG algorithm, which groups the original gray-level bins into a default number of bins, 20, executes much faster than the optimized
GLG, and has comparable results.

repeat the above procedure and group the histogram com- imal will lead to the corresponding optimum number
ponents into all possible numbers from to 2 (there is for gray-level bins
no need to group all histogram components into one bin
since the histogram will be the same as the original after it
for (17)
is ungrouped), and calculate the average distance between
pixels on the grayscale, , for each set of bins. The max- for (18)
2298
Fig. 9. Comparison of GLG results using different gray-level bin numbers.
Both images are satisfactory. (a) GLG result of the Phobos image with the op-
timal bin number of 4, found through the iterative process. (b) FGLG result of
the Phobos image with the default bin number of 20.
In our example, we continue to group the gray-level bins.
This time bin and will be grouped as in-
dicated by the bracket in Fig. 6(b), and the new set of
gray-level bins are
Their boundaries are
Then and the new transformed gray
levels are
The resulting histogram is shown in Fig. 6(c). The average
distance between pixels on the grayscale is
larger than and . result. This effect also exists in the results of HE and histogram
• To obtain the final optimally enhanced image, we retrieve
the optimal gray-level transformation function
from computer memory, and then apply it to the original
image.
Fig. 7 illustrates the flow chart of the optimized gray-level
grouping algorithm. Figs. 1(d) and 2(d) show the result of ap-
plying this technique to the Phobos image and the resulting
histogram, respectively. Fig. 4(c) shows the GLG result of the
X-ray luggage image. Since the background area percentage
is quite large in the X-ray image, the background is excluded
IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 8, AUGUST 2006
Fig. 10. Mars moon—phobos. (a) Low-contrast original image. (b) Result of
CLAHE. (c) Result of global gray-level grouping. (d) Result of A-GLG. Its con-
trast enhancement is the strongest (original image courtesy of Dr. R. C. Gonzalez
[1]).
when calculating the average distance between pixels on the
grayscale, in order to achieve the strongest enhancement in the
resulting image of Fig. 4(c). It is obvious that the GLG results
are better than those of histogram equalization and histogram
specification. Furthermore, the new method is fully automatic.
It should be pointed out that in Fig. 1(d), background noise
in the upper left corner has been amplified in the basic GLG
specification, as shown in Fig. 1(b) and (c). This problem can
be solved and will be addressed in Part II of this paper.
Although we have described an approach for finding the op-
timal number of gray-level groups, it has been found that the
quality of the resulting images is not very sensitive to the total
number of gray-level bins if this number is below 150, because
the ungrouping of grouped gray levels results in similar spacing
between high-amplitude histogram components. Therefore, we
can use a default value for the total number of gray-level groups,
e.g., 20, which has been found to yield satisfactory results in a
CHEN et al.: GLG: AN AUTOMATIC METHOD FOR OPTIMIZED IMAGE CONTRAST ENHANCEMENT—PART I 2299

Fig. 11. X-ray image of luggage. (a) Original image. (b) Result of CLAHE. (c) Result of global GLG. (d) Result of A-GLG. Not only its contrast enhancement
is the strongest, but also some objects that are invisible in the original image and the CLAHE result become readily apparent (e.g., the tip of the stripe to the right
of the luggage, as shown by the white oval in the image). (The Tenengrad value of the global GLG result for this image is a little higher than that of the adaptive
GLG result, because some light regions in conjunction with the surrounding dark background in the global GLG result exhibit larger gradient magnitudes, which
lead to a larger Tenengrad value, but the adaptive GLG result apparently has a higher local contrast enhancement).

large number of experiments and saves on iterations and com-
putational expenses significantly. The choice of this number is
also based on the fact that it is comparable to the number of
gray shades that the human eye can discern, which is a couple
of dozens. Without constructing the transformation function and
calculating the average distance between pixels on the grayscale
for each set of gray-level bins, this algorithm executes much
faster (more than 3 orders of magnitude faster for 8-bit images)
than the optimized GLG, so it is called fast gray-level grouping
(FGLG). Fig. 8 illustrates the flow chart of FGLG.
Fig. 9(a) and (b) shows the comparison of the results of
treating the Phobos image by GLG with two different numbers
of gray-level bins. Fig. 9(a) is the result using the optimal
number of bins of four given by (18), and Fig. 9(b) is the result
using the default number of bins of 20. It can be seen that there
is not much difference in the two images, and both images are
satisfactory.
III. COMPUTATIONAL COMPLEXITY AND QUALITY MEASURE
The computational complexity of the GLG technique is ba-
sically determined by the number of comparison operations for
finding the smallest gray-level groups and the number of mul-
tiplication and/or division operations for calculating the gray-
level transformation functions . kernels and . The gradient magnitude is given as
The number of comparison operations for finding the smallest
gray-level group among groups is . Since the smallest
gray-level group needs to be found for all possible sets of groups
in the optimal GLG process, the computational complexity for
finding the smallest gray-level groups is , where is the
total number of gray levels on the grayscale. For each set of
gray-level groups, the number of multiplication and/or division
operations for calculating the gray-level transformation function
is . Since this also needs to be performed on all possible
sets of gray-level groups in the optimal GLG process, the com-
putational complexity for calculating gray-level transformation
functions in the optimal GLG process is . However, since
the gray-level transformation function is calculated only once in
the fast GLG process, its computational complexity for multi-
plication operations is . As comparison, the computational
complexity of the HE algorithm is .
In order to evaluate the competitiveness of the GLG method
against existing contrast enhancement techniques, we used
the most well-known benchmark image sharpness measure,
the Tenengrad criterion [26], [27], to compare the results of
the GLG method and the conventional methods studied in this
paper. The Tenengrad criterion is based on gradient magnitude
maximization. It is considered one of the most robust and func-
tionally accurate image quality measures [27]. The Tenengrad
value of an image is calculated from the gradient at
each pixel , where the partial derivatives are obtained by
a high-pass filter, e.g., the Sobel operator, with the convolution
(19)
2300 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 8, AUGUST 2006

and the Tenengrad criterion is formulated as b) For the boundary components

for
for (20) for
otherwise
(22)
where is a threshold. The image quality is usually considered
for , , and
higher if its Tenengrad value is larger. , and
We calculated the Tenengrad values of all images in
this article, and listed them in the corresponding figure captions.
It is noted that the images processed with the GLG technique for
have significantly larger Tenengrad values, which indicates that for
the GLG method is superior to the conventional techniques com- otherwise
pared to in this paper. This result agrees with the visual evalua- (23)
tion by the human eye. It is also worth noting that the Tenengrad for , , and
.
criterion indicates that the optimal GLG result is better than the
c) For the interior components
fast GLG result, as shown in Fig. 9.
In the previous section, we proposed an image contrast mea-
sure, the average distance between pixels on the grayscale. We for (24)
also calculated the values of this criterion for all im-
ages in this article and listed them in the corresponding figure
captions. It can be seen that, this criterion generally agrees well where , , and ,
with the benchmark Tenengrad measure in evaluating image 3, 2, or 1, which equals to the number of operands
contrast. It also should be noted that, the values of in the numerator. The above equation applies to
, , and
the adaptive GLG (A-GLG) results described in Section IV usu-
.
ally do not agree with the perceived image contrasts, because
This step is an averaging process to balance the contrast
the adaptive GLG process significantly changes the global his-
of adjacent subimages in the final output image. If gray-
togram profile of the image, and therefore makes the comparison
level in the original image is mapped to gray-level
of the values of the global GLG and adaptive GLG by , it is considered as background and therefore
results meaningless. This is one of the situations in which the excluded from the averaging process.
criterion should not be used. 3) Perform bilinear interpolation to reconstruct the final
output image. For each original subimage , func-
IV. ADAPTIVE GRAY-LEVEL GROUPING (A-GLG) tion returns the gray-level value of the
pixel at subimage coordinate, , for ,
By analogy to the well-established conventional technique of , where and are the height and
AHE [4], or contrast-limited adaptive histogram equalization width of the corresponding subimage, respectively. The
(CLAHE), GLG also has its adaptive counterparts—A-GLG, or bilinearly-interpolated output subimage is given
CLA-GLG. In the A-GLG or CLA-GLG method, the image is by the following expression:
first divided into an array of subimages (usually an 8 8 array),
each subimage is treated with the GLG method, and all treated
subimages are merged together by bilinear interpolation to gen-
erate the processed whole image.
The algorithm of adaptive GLG (A-GLG) technique is de-
scribed as the following.
1) Divide the original image into an array of subim-
ages, and process all subimages with the GLG algorithm to (25)
obtain their optimal GLG gray-level transformation func- for , , and
tions, i.e., , for , , .
and . Here, represents the highest The final processed whole image is obtained by stitching
gray-level value on the grayscale. the array of output subimages together.
2) Create an intermediate array Fig. 10(c) shows the A-GLG result of the Phobos image with
of gray-level transformation functions for comparison to the CLAHE result and global GLG result. It can
, , and be seen that the A-GLG result is bviously better than CLAHE
, as follows. and global GLG results.
a) For the four corner components AHE is often not applicable to many images such as images
with large areas of dark background. In that case, because of
its adaptive nature, AHE will turn the dark background into
white, and cause undesirable artifacts. Therefore, CLAHE is
(21) more applicable than AHE. However, A-GLG does not have
CHEN et al.: GLG: AN AUTOMATIC METHOD FOR OPTIMIZED IMAGE CONTRAST ENHANCEMENT—PART I
this problem, and will automatically keep the dark background
black.
Fig. 11(c) shows the A-GLG result of the X-ray image of
luggage with comparison to the CLAHE result and global GLG
result. It can be seen that not only the contrast of the A-GLG
result is better than that of the CLAHE result (e.g., the razor in
the bag is brighter and clearer), but also some objects that are
invisible in the original image and the CLAHE result become
readily apparent (e.g., the tip of the stripe to the right of the
luggage, as shown by the white oval in the image).
V. CONCLUSION
We have developed a new automatic contrast enhancement
technique. GLG is a general and powerful technique, which
can be conveniently applied to a broad variety of low-contrast
images and generates satisfactory results (more examples will
be given in Part II of this paper). The GLG technique can be
conducted with full automation at fast speeds and outperforms
conventional contrast enhancement techniques. The benchmark
image quality measure, Tenengrad criterion, indicates that the
GLG results are superior to those of conventional techniques
studied in this paper. The optimized GLG algorithm generally
can process an image within a few seconds on a personal com-
puter (PC), and the FGLG algorithm can process an image on
the time scale of millisecond on a PC. The basic GLG method
also provides a platform for various extensions of this technique,
such as selective gray-level grouping (SGLG), (S)GLG with
preprocessing steps for eliminating image background noises,
(S)GLG on color images, and so on. All these variations extend
the capability of the basic GLG technique, and will be discussed
in Part II of this paper.
ACKNOWLEDGMENT
The authors would like to thank Y. Yao for help with the
Tenengrad criterion, and Dr. A. Koschan and Dr. A. Gribok for
their insightful and valuable suggestions.
REFERENCES
[1] R. C. Gonzalez and R. E. Woods, Digital Image Processing, 2nd ed.
Englewood Cliffs, NJ: Prentice-Hall, 2002, ISBN: 0-201-18075-8.
[2] Y.-T. Kim, “Enhancement using brightness preserving bi-histogram
equalization,” IEEE Trans. Consumer Electron., vol. 43, no. 1, pp. 1–8,
Feb. 1997.
[3] Y. Wan, Q. Chen, and B.-M. Zhang, “Image enhancement based on
equal area dualistic sub-image histogram equalization method,” IEEE
Trans Consumer Electron., vol. 45, no. 1, pp. 68–75, Feb. 1999.
[4] S. M. Pizer, E. P. Amburn, J. D. Austin, R. Cromartie, A. Geselowitz, T.
Greer, B. H. Romeny, J. B. Zimmerman, and K. Zuiderveld, “Adaptive
histogram equalization and its variations,” Comput. Vis., Graph., Image
Process., vol. 39, pp. 355–368, 1987.
[5] T.-K. Kim, J.-K. Paik, and B.-S. Kang, “Contrast enhancement
system using spatially adaptive histogram equalization with temporal
filtering,” IEEE Trans. Consumer Electron., vol. 44, no. 1, pp. 82–86,
Feb. 1998.
[6] V. Caselles, J.-L. Lisani, J.-M. Morel, and G. Sapiro, “Shape preserving
local histogram modification,” IEEE Trans. Image Process., vol. 8, no.
2, pp. 220–230, Feb. 1999.
[7] D.-C. Chang and W.-R. Wu, “Image contrast enhancement based on
a histogram transformation of local standard deviation,” IEEE Trans.
Med. Imag., vol. 17, no. 4, pp. 518–531, Aug. 1998.
[8] S.-D. Chen and A. R. Ramli, “Contrast enhancement using recursive
mean-separate histogram equalization for scalable brightness preserva-
tion,” IEEE Trans. Consumer Electron., vol. 49, no. 4, pp. 1301–1309,
Nov. 2003.
2301
[9] ——, “Minimum mean brightness error bi-histogram equalization in
contrast enhancement,” IEEE Trans. Consumer Electron., vol. 49, no.
4, pp. 1310–1319, Nov. 2003.
[10] S. Dippel, M. Stahl, R. Wiemker, and T. Blaffert, “Multiscale contrast
enhancement for radiographies: laplacian pyramid versus fast wavelet
transform,” IEEE Trans. Med. Imag., vol. 21, no. 4, pp. 343–353, Apr.
2002.
[11] Y. Jin, L. Fayad, and A. Laine, “Contrast enhancement by multi-scale
adaptive histogram equalization,” in Proc. SPIE, 2001, vol. 4478, pp.
206–213.
[12] J.-Y. Kim, L.-S. Kim, and S.-H. Hwang, “An advanced contrast en-
hancement using partially overlapped sub-block histogram equaliza-
tion,” IEEE Trans. Circuits Syst. Video Technol., vol. 11, no. 4, pp.
475–484, Apr. 2001.
[13] S.-Y. Kim, D. Han, S.-J. Choi, and J.-S. Park, “Image contrast en-
hancement based on the piecewise-linear approximation of CDF,”
IEEE Trans. Consumer Electron., vol. 45, no. 3, pp. 828–834, Aug.
1999.
[14] T.-K. Kim, J.-K. Paik, and B.-S. Kang, “Contrast enhancement
system using spatially adaptive histogram equalization with temporal
filtering,” IEEE Trans. Consumer Electron., vol. 44, no. 1, pp. 82–87,
Feb. 1998.
[15] S. C. Matz and R. J. P. de Figueiredo, “A nonlinear technique for image
contrast enhancement and sharpening,” in Proc. IEEE Int. Symp. Cir-
cuits and Systems, 1999, vol. 4, pp. 175–178.
[16] S. K. Naik and C. A. Murthy, “Hue-preserving color image enhance-
ment without gamut problem,” IEEE Trans. Image Process., vol. 12,
no. 12, pp. 1591–1598, Dec. 2003.
[17] J. P. Oakley and B. L. Satherley, “Improving image quality in poor
visibility conditions using a physical model for contrast degradation,”
IEEE Trans. Image Process., vol. 7, no. 2, pp. 167–179, Feb. 1998.
[18] S.-C. Pei, Y.-C. Zeng, and C.-H. Chang, “Virtual restoration of ancient
chinese paintings using color contrast enhancement and lacuna texture
synthesis,” IEEE Trans. Image Process., vol. 13, no. 3, pp. 416–429,
Mar. 2004.
[19] A. Polesel, G. Ramponi, and V. J. Mathews, “Image enhancement via
adaptive unsharp masking,” IEEE Trans. Image Process., vol. 9, no. 3,
pp. 505–510, Mar. 2000.
[20] F. Russo, “An image enhancement technique combining sharpening
and noise reduction,” IEEE Trans. Instrum. Meas., vol. 51, no. 4, pp.
824–828, Aug. 2002.
[21] F. Sattar and D. B. H. Tay, “Enhancement of document images using
multiresolution and fuzzy logic techniques,” IEEE Signal Process.
Lett., vol. 6, no. 10, pp. 249–252, Oct. 1999.
[22] J.-L. Starck, F. Murtagh, E. J. Candes, and D. L. Donoho, “Gray and
color image contrast enhancement by the curvelet transform,” IEEE
Trans. Image Process., vol. 12, no. 6, pp. 706–717, Jun. 2003.
[23] J. A. Stark, “Adaptive image contrast enhancement using generaliza-
tions of histogram equalization,” IEEE Trans. Image Process., vol. 9,
no. 5, pp. 889–896, May 2000.
[24] J. Yang, G. Dong, Y. Peng, Y. Yamaguchi, and H. Yamada, “General-
ized optimization of polarimetric contrast enhancement,” IEEE Geosci.
Remote Sens. Lett., vol. 1, no. 3, pp. 171–174, Jul. 2004.
[25] X. Zong, A. F. Laine, and E. A. Geiser, “Speckle reduction and contrast
enhancement of echocardiograms via multiscale nonlinear processing,”
IEEE Trans. Med. Imag., vol. 17, no. 4, pp. 532–540, Aug. 1998.
[26] E. P. Krotkov, Active Computer Vision by Cooperative Focus and
Stereo. New York: Springer-Verlag, 1989.
[27] A. Buerkle, F. Schmoeckel, M. Kiefer, B. P. Amavasai, F. Caparrelli,
A. N. Selvan, and J. R. Travis, “Vision-based closed-loop control of
mobile microrobots for micro handling tasks,” Proc. SPIE, vol. 4568,
Microrobotics and Microassembly III, pp. 187–198, 2001.
ZhiYu Chen (SM’98) received the B.E. degree (with
high honors) in engineering physics and the M.S.
degree from Tsinghua University, Beijing, China,
in 1994 and 1997, respectively. He is currently
pursuing the Ph.D. degree in electrical engineering
at the University of Tennessee, Knoxville (UTK).
His area of research was plasma science and tech-
nology. He has authored or coauthored seven journal
articles and numerous conference papers in that field.
In late 2003, he switched his research interest to the
field of digital image processing and computer vision.
He is now with the Imaging, Robotics and Intelligent Systems (IRIS) Labora-
tory at UTK.
Mr. Chen was a recipient of the IEEE Nuclear and Plasma Sciences So-
ciety Graduate Student Award in 2002. He also serves as a Reviewer for IEEE
TRANSACTIONS ON PLASMA SCIENCE.
2302 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 8, AUGUST 2006

Besma R. Abidi (M’88–SM’06) received the M.S.
degrees in image processing and remote sensing
(Hons.) from the National Engineering School of
Tunis, Tunis, Tunisia, in 1985 and 1986, respec-
tively, and the Ph.D. degree from The University of
Tennessee, Knoxville (UTK), in 1995.
She is a Research Assistant Professor with the
Department of Electrical and Computer Engineering,
UTK, which she joined in 1998. She occupied a
postdoctorate position with the Oak Ridge Institute
of Science and Energy, Oak Ridge, TN, and was a
Research Scientist at the Oak Ridge National Laboratory from 1998 to 2001.
From 1985 to 1988, she was an Assistant Professor at the National Engineering
School of Tunis. Her general areas of research are in image enhancement and
restoration, sensor positioning and geometry, video tracking, sensor fusion,
and biometrics.
Dr. Abidi is a member of SPIE, Tau Beta Pi, Eta Kappa Nu, Phi Kappa Phi,
and The Order of the Engineer.
Mongi A. Abidi (M’82) received the M.S. and Ph.D.
degrees in electrical engineering from the University
of Tennessee, Knoxville (UTK), in 1985 and 1987,
respectively.
He is a W. Fulton Professor with the Department
of Electrical and Computer Engineering, UTK,
which he joined in 1986. His interests include image
processing, multisensor processing, three-dimen-
sional imaging, and robotics. He has published over
120 papers in these areas and co-edited the book
Data Fusion in Robotics and Machine Intelligence
(Academic, 1992).
Dr. Abidi received the 1994–1995 Chancellor’s Award for Excellence in Re-
search and Creative Achievement and the 2001 Brooks Distinguished Professor
Award. He is a member of the Computer Society, Pattern Recognition Society,
SPIE, Tau Beta Pi, Phi Kappa Phi, and Eta Kappa Nu honor societies. He also
received the First Presidential Principal Engineer Award prior to joining the Uni-
versity of Tennessee.

David L. Page (M’91) received the B.Sc. and M.Sc.

degrees from Tennessee Technological University,
Cookeville, in 1993 and 1995, respectively, and
the Ph.D. degree from the University of Tennessee,
Knoxville (UTK), in 2003, all in electrical engi-
neering.
He then began work as an Electronics Engineer
at the Naval Surface Warfare Center, Dahlgren, VA,
where he was involved in distributed computing re-
search. He currently serves as a Research Assistant
Professor in the Imaging, Robotics, and Intelligent
Systems Laboratory at UTK. His research interests focus on three-dimensional
computer vision, with specific emphasis on mesh segmentation, curvature esti-
mation, and object description.